B&H Photo is a paid partner of Adam Talks Tech. Some of the links in this post are affiliate links, which means I may earn a small commission if you choose to make a purchase at no extra cost to you. As an Amazon Associate and a B&H affiliate, I earn from qualifying purchases. Read my full Ethics & Affiliate Relationship Disclaimer.
There are two versions of the M5 Pro Mac mini. The base one has a 15-core CPU and a 16-core GPU, and for $200 more you can get an 18-core CPU with a 20-core GPU. Same memory, same storage, same ports. So is that upgrade worth it?
I ran both of them through my full set of tests: Final Cut Pro, Lightroom Classic, Xcode, Logic Pro, Blender, gaming, local AI, Geekbench 7 and Puget Bench. Big thanks to B&H for sending both of these over to test. Every chart from the video is below, so you can see exactly where those extra cores show up and where they don’t.
If you write code, make music or do 3D, get the 18-core. Those were 15 to 21% faster. For video and photo work the two were basically tied, so I’d keep the $200 and put it toward more RAM.
- Xcode builds about 21% faster, 87 seconds vs 105
- Logic played 378 tracks vs 328 and bounced in 1:26 vs 1:40
- Blender CPU render in 2:53 vs 3:19
- About 24% higher Geekbench 7 multi-core score
- $200 less, and the base one goes on sale a lot
- Final Cut exports were a tie, 1:19 on both
- Lightroom exports and AI Denoise were basically the same
- Puget Bench for DaVinci Resolve and Premiere Pro only went up about 5% on the 18-core
Specs and price
The base M5 Pro Mac mini is $1,699. You get a 15-core CPU, a 16-core GPU, 24GB of memory and 512GB of storage. The upgraded one has the 18-core CPU and 20-core GPU with the same 24GB and 512GB for $1,899. That’s $200 more, about 12%. Heads up, the base one goes on sale a lot. It was $1,549 at B&H when I made this video.
| M5 Pro Mac mini, 15-core | M5 Pro Mac mini, 18-core | |
|---|---|---|
| Price | $1,699 | $1,899 |
| CPU | 15-core (5 super, 10 performance) | 18-core (6 super, 12 performance) |
| GPU | 16-core | 20-core |
| Memory | 24GB | 24GB |
| Storage | 512GB | 512GB |
| Memory bandwidth | 307GB/s | 307GB/s |
| Ports | Same | Same |
If you want help picking a Mac without reading this whole thing, I offer free personalized Mac buying recommendations. Tell me how you use your Mac and I’ll get back to you with a pick.
It’s the same chip
Here’s something most people don’t realize. These are the same chip. When Apple makes M5 Pro chips, not every one comes out perfect. If a few cores don’t pass testing, they get switched off, and that chip becomes the base model. That’s called binning.
So the base one is the exact same chip with 3 CPU cores and 4 GPU cores turned off. It has the same memory speed and the same ports.
Real-world tests
Every chart below pulls straight from my Mac test results, so the numbers here match what I show in the video. Both Macs were on the same version of macOS.
Final Cut Pro
I started with Final Cut, since that’s what I edit in every day. A 3-minute 4K video with the social media 4K setting took 1:19 on both. The exact same time. A 30-minute 4K video with the Apple Devices 4K setting took 12:51 on the base and 12:53 on the upgraded. So for exports it’s a tie. Those extra cores just don’t show up here.
Magnetic Mask is the AI tool that cuts a person out of a clip. On an 18-second clip it took 23.9 seconds on the base and 24.7 on the upgraded. Tie again. The one place Final Cut showed a difference was playback. I stacked 6K clips to see how many each Mac could play at the same time, and the base played 12 while the upgraded played 14. So if you cut multicam with a lot of 6K angles, that’s where the upgrade helps.
| Final Cut Pro, 30-minute 4K export | M5 Pro Mac mini, 15-core | M5 Pro Mac mini, 18-core |
|---|---|---|
| Apple Devices 4K setting (lower is better) | 12:51 | 12:53 |
Lightroom Classic
When you import photos, Lightroom builds previews so you can scroll through them fast. That was faster on the upgraded, 16.8 seconds vs 20.3. But exporting 500 photos as full-quality JPEGs was a tie, 1:31 on the base and 1:35 on the upgraded. AI Denoise on 50 RAW photos took 5:53 vs 6:02, basically the same. So if you’re mostly editing photos, I’d keep the $200.
Xcode
Now here’s where the 18-core starts to earn it. I built the same Xcode project three times on each Mac and took the average. 105 seconds on the base, 87 on the upgraded. That’s about 21% faster, and if you’re building code all day, that time adds up.
| Xcode benchmark build | M5 Pro Mac mini, 15-core | M5 Pro Mac mini, 18-core |
|---|---|---|
| Build time, average of 3 runs (lower is better) | 105.07 sec | 86.66 sec |
Logic Pro
I kept adding tracks until each Mac maxed out. The base played 328 tracks and the upgraded played 378. Then I bounced a 512-track project. A bounce is when Logic exports every track down into one audio file. That took 1:40 on the base and 1:26 on the upgraded.
Blender
I rendered the Classroom scene on the CPU, 3:19 vs 2:53. The heavier Barbershop scene on the GPU took 1:50 vs 1:42. One frame doesn’t tell you much, so I also rendered an animation: the first 25 frames of the Red Autumn Forest demo scene on the GPU. That took 6:44 on the base and 6:03 on the upgraded, about 11% faster. For code, music and 3D, you can actually feel those extra cores.
| Blender 5.2 animation, GPU | M5 Pro Mac mini, 15-core | M5 Pro Mac mini, 18-core |
|---|---|---|
| Red Autumn Forest, frames 1-25, Cycles, 128 samples (lower is better) | 6:44 | 6:03 |
Gaming
I ran the built-in Cyberpunk 2077 benchmark at 1080p on High. With no upscaling it was 60 frames per second on the base and 68 on the upgraded. With MetalFX turned on, that’s Apple’s upscaling, it was 89 vs 94. Frame generation adds in-between frames so it looks smoother, and that got 138 vs 151. With ray tracing on, both slow way down, 39 vs 43.
I also tested Minecraft, and it runs great on both. At render distance 16 it was about 125 frames per second on the base and 143 on the upgraded. Push it to 32 and they tie at about 83. The upgrade gets you about 6 to 12% more frames, but both play Cyberpunk at over 60 fps at 1080p High, so I wouldn’t pay $200 just for gaming.
| Frames per second (higher is better) | M5 Pro Mac mini, 15-core | M5 Pro Mac mini, 18-core |
|---|---|---|
| Cyberpunk 2077, 1080p High, native | 60.12 | 67.62 |
| Cyberpunk 2077, MetalFX Quality | 89.19 | 94.20 |
| Cyberpunk 2077, frame generation | 138.25 | 150.97 |
| Cyberpunk 2077, ray tracing | 39.26 | 43.44 |
| Minecraft, 1080p Fancy, render distance 16 | about 125 | about 143 |
| Minecraft, 1080p Fancy, render distance 32 | about 83 | about 83 |
Gaming isn’t part of my Mac test results sheet yet, so these are in a table instead of a chart. Cyberpunk 2077 2.5 built-in benchmark, VSync off, same monitor on both.
Local AI
In Diffusion Bee I had both make the same image, a cat working on a laptop with nachos on the side. 8.7 seconds on the base, 7.5 on the upgraded. That’s about 17% faster. In LM Studio I had it write 50 paragraphs about the Roman Empire. The base ran about 92 tokens per second and the upgraded ran about 98, only about 6%. Both have the same memory, so LM Studio mostly comes down to memory bandwidth, and that’s the same on both.
Benchmarks
Geekbench 7
Single-core was about the same, 3,513 vs 3,558. That makes sense, it’s the same chip. Multi-core is where those extra CPU cores show up, 26,454 vs 32,851, about 24% higher. On the GPU, the Metal score was 128,469 vs 138,874, about 8% higher.
Puget Bench
I also ran Puget Bench. It’s from Puget Systems, and it runs real tasks inside the actual apps. I don’t edit in Premiere, After Effects or Resolve myself, so this let me cover those apps too.
| Puget Bench overall score (higher is better) | M5 Pro Mac mini, 15-core | M5 Pro Mac mini, 18-core |
|---|---|---|
| Puget Bench for DaVinci Resolve (Standard) | 71,868 | 75,111 |
| Puget Bench for Premiere Pro (Extended) | 80,961 | 85,270 |
| Puget Bench for After Effects (Standard) | 11,216 | 11,222 |
| Puget Bench for Lightroom Classic (Standard) | 10,400 | 10,489 |
| Puget Bench for Photoshop (Standard) | 16,825 | 17,353 |
Puget Bench for DaVinci Resolve and Puget Bench for Premiere Pro were both about 5% higher on the upgraded. Puget Bench for After Effects, Lightroom Classic and Photoshop were basically ties. Puget Bench for Lightroom Classic did build smart previews about 11% faster on the upgraded, same as my own Lightroom test, but exports were a tie.
Inside those scores, a few tasks jumped a lot more. In Puget Bench for DaVinci Resolve, DNxHR playback was about 21% faster, ProRes 422 HQ exports were about 18% faster, and RED and Blackmagic RAW were about 15% faster. In Puget Bench for Premiere Pro, DNxHR exports were 15 to 18% faster and the heavier GPU effects were 10 to 20% faster. In Puget Bench for After Effects, 3D renders were about 17 to 22% faster. But H.264 exports, the format most people use for YouTube, were a tie in Puget Bench for DaVinci Resolve and Premiere Pro. So if you work in ProRes, DNxHR or RAW, the upgrade helps. If you export H.264 for YouTube, it doesn’t.
| Puget Bench task (frames per second, higher is better) | M5 Pro Mac mini, 15-core | M5 Pro Mac mini, 18-core |
|---|---|---|
| Puget Bench for DaVinci Resolve: 4K DNxHR LB playback | 252.9 | 306.7 (+21%) |
| Puget Bench for DaVinci Resolve: ProRes 422 HQ UHD export | 266.8 | 314.4 (+18%) |
| Puget Bench for DaVinci Resolve: 4K RED R3D playback | 66.4 | 77.1 (+16%) |
| Puget Bench for DaVinci Resolve: 4.6K Blackmagic RAW playback | 126.1 | 144.9 (+15%) |
| Puget Bench for Premiere Pro: DNxHR LB UHD export | 89.3 | 104.5 (+17%) |
| Puget Bench for Premiere Pro: DNxHR SQ UHD export | 92.2 | 106.4 (+15%) |
| Puget Bench for Premiere Pro: DNxHR HQX UHD export | 55.8 | 65.7 (+18%) |
| Puget Bench for Premiere Pro: VR De-Noise x4 (GPU effect) | 4.27 | 5.12 (+20%) |
| Puget Bench for Premiere Pro: VR Digital Glitch x20 (GPU effect) | 13.2 | 15.7 (+19%) |
| Puget Bench for Premiere Pro: Bokeh Blur x2 (GPU effect) | 4.67 | 5.17 (+11%) |
| Puget Bench for After Effects: 3D Simple Text render | 4.07 | 4.78 (+17%) |
| Puget Bench for After Effects: 3D Depth render | 1.51 | 1.84 (+22%) |
| Puget Bench for DaVinci Resolve: H.264 50Mbps UHD export | 58.0 | 57.6 (tie) |
| Puget Bench for Premiere Pro: H.264 50Mbps UHD export | 71.5 | 73.5 (tie) |
Puget Bench results from Puget Systems: pugetsystems.com/pugetbench/creators. Single runs on each Mac on macOS 27.2. Puget Bench for DaVinci Resolve 2.1.1 (Resolve Studio 21.2), Puget Bench for Premiere Pro 2.1.0 beta (Premiere 27.0), Puget Bench for After Effects 1.2.0 beta (After Effects 27.0), Puget Bench for Lightroom Classic 1.0.0 (Lightroom Classic 15.2), Puget Bench for Photoshop 1.0.7 (Photoshop 28.0).
Is it 12% faster?
The upgrade costs about 12% more, so the question is whether you get at least 12% more speed. In Xcode, Logic and Blender on the CPU, yes. Those were 15 to 21% faster, and Diffusion Bee was 17% faster too. In Final Cut, Lightroom and Photoshop, no. Those were a tie or a hair slower. Gaming landed around 6 to 12%, close to what you’re paying but not more. LM Studio only went up about 6%, and Puget Bench for DaVinci Resolve and Premiere Pro only went up about 5%.
| Yes, 12% or more | No, under 12% |
|---|---|
| Xcode build +21% | Blender animation +11% |
| Diffusion Bee +17% | Blender GPU +8% |
| Logic bounce +16% | LM Studio +6% |
| Logic tracks +15% | Puget Bench for Premiere Pro and DaVinci Resolve +5% |
| Blender CPU +15% | Final Cut and Lightroom exports: tie |
| Cyberpunk +12% |
Which one should you buy?
If you write code, make music or do 3D, get the upgrade. If that’s not you, pass on it. And if you can only get one upgrade, I’d probably pick the RAM for performance. Going from 24GB to 48GB is $600 more.
| Get the 15-core if | Get the 18-core if |
|---|---|
| You mostly edit video in Final Cut or export H.264 for YouTube | You build code in Xcode all day |
| You mostly edit photos in Lightroom or Photoshop | You make music in Logic Pro |
| You want a Mac that games well, both run Cyberpunk over 60 fps | You render in Blender on the CPU |
| You’d rather put the $200 toward more RAM | You work in ProRes, DNxHR or RAW |
Buying from B&H? Their Payboo card can save you the sales tax.
The bottom line
The 18-core is the same chip with more of it turned on. It pays off when you push the CPU. For everything else, the base M5 Pro Mac mini is plenty, and I’d put that $200 toward more RAM.
Every result from every Mac I’ve ever tested is on my Mac test results page. And if you’re still stuck on which one to get, tell me how you use your Mac and I’ll help you pick.
Still deciding? If you’re between the M5 Pro mini and the cheaper M6, see my M6 Mac mini vs M5 Pro Mac mini tests. And if you need more than an M5 Pro, here’s the M5 Max Mac Studio vs M5 Pro Mac mini.
